Digital News

　　At the 2025 World Artificial Intelligence Conference (WAIC 2025), Huawei’s Atlas 900 A3 SuperPoD attracted a steady stream of visitors, drawing widespread attention for its impressive 300 PFlops of computing power; INESA’s Intelligent Computing Power Scheduling Platform 2.0 connected with 15 providers, making over 18,000 PFlops of computing power available for enterprise collaboration and negotiations; meanwhile, Enflame’s booth showcasing its intelligent computing deployments and inference card applications became a hotspot for lively technical discussions… These vivid scenes are a microcosm of China’s rapid growth in the computing power industry.

　　Computing power serves as the foundation of the digital economy and the core engine driving new quality productive forces. It is not only essential for enabling the transformation and upgrading of industries across the board but also a strategic force reshaping the global competitive landscape. In the first half of this year, amid intensifying global competition in the digital economy and accelerated technological advancements, China’s computing power industry has firmly committed to a path of high-quality development. Strategically guided by the “East Data, West Computing” initiative, it has continuously strengthened its computing infrastructure and accelerated the construction of a national integrated computing power network. A high-quality industrial landscape, characterized by intelligent computing leadership, computing power network collaboration, and green, low-carbon development, is gradually taking shape, laying a solid foundation for the growth of new quality productive forces. 

　　A Surge in Intelligent Computing Capacity

　　Powering Large-Scale AI Applications

　　In Hunan’s Xiangjiang New Area, newly deployed “AI civil servants” help the public by answering questions and instantly transforming lengthy policy documents into clear, concise summaries. Meanwhile, in the server room of Tianjin Medical University General Hospital, the DeepSeek “Intelligent Computing Integrated Machine” supports comprehensive evaluations of specialized chronic diseases in the elderly and generates AI-driven health examination reports… The rapid explosion of artificial intelligence (AI) has driven an unprecedented demand for computing power. In particular, intelligent computing has become the essential backbone supporting the stable operation of next-generation AI technologies, represented by large models and intelligent agents.

　　According to the latest data from the Ministry of Industry and Information Technology, China’s intelligent computing capacity reached 748 EFlops in the first quarter of this year, accounting for 35% of the country’s total computing power and emerging as the primary driver of growth in the sector. International Data Corporation (IDC) forecasts that by 2025, China’s intelligent computing capacity could rise to 1,037.3 EFlops, a year-on-year increase of over 40%, solidifying its position as a global leader.

　　The surge in intelligent computing capacity is backed by the ongoing capacity expansion and quality improvement of China’s computing infrastructure. Data shows that currently, China has 10.43 million standard computing racks in use, an 18.5% increase over 2024.

　　To better support the development and deployment of large models and other AI technologies and applications, China Telecom’s Central Intelligent Computing Center in Wuhan, Hubei, recently expanded its computing capacity by adding 300 intelligent computing servers, boosting its computing power by 2 EFlops. This upgrade aims to more effectively meet the AI needs of organizations such as the Wuhan AI Research, Transn IOL Technology, and Wuhan Zhongyuan Electronics. At the Qingyang data center cluster, part of the Gansu hub of the “East Data, West Computing” initiative, intelligent computing capacity has exceeded 50 EFlops, with 31,000 standard racks installed. The cluster primarily serves computing demands from major regions, including the Beijing-Tianjin-Hebei region, the Yangtze River Delta, and the Guangdong-Hong Kong-Macao Greater Bay Area. In addition, telecom operators have built intelligent computing clusters with up to 10,000 GPUs, creating a “super hub” for computing power supply—capable of supporting the training of trillion-parameter models and large-scale data processing…

　　According to data from the National Data Administration, by the first quarter of this year, the total computing power of the eight major hubs of China’s “East Data, West Computing” initiative had reached 215.5 EFlops, with intelligent computing power accounting for 80.8% of that total. This reflects China’s strong momentum in expanding both the scale and quality of its computing power industry.

　　Computing Power Network Collaboration

　　Building a National Integrated Framework

　　While building a robust computing power network is essential, ensuring it is accessible and easy to use is even more critical. The vision is to pool computing power resources nationwide into a shared “reservoir,” allowing anyone to draw what they need, whenever they need it. This is key to promoting inclusive access to computing power.

　　“Our manufacturing cycle has been shortened from 60 months to 24 months.” In Hangzhou, Zhejiang, the “1ms All-Optical Intelligent Computing Dedicated Network” enables Leapmotor to efficiently allocate its computing power resources. Zhang San, Senior Expert at Leapmotor’s Security Development Department, told the reporter that by leveraging the computing network collaboration, Leapmotor has achieved intelligent upgrades throughout the entire process from R&D to production, significantly shortening its design iteration cycle.

　　In the first half of the year, the Ministry of Industry and Information Technology, along with other agencies, implemented targeted measures by officially launching the construction of a computing power Internet test network and introducing policies such as the Computing Power Interconnection Action Plan. Following a systematic approach based on the “points, chains, networks, and the totality,” they advanced computing power-related work in an organized manner. Significant progress has been made in building a national integrated computing power network, with comprehensive supply capabilities steadily improving, enabling over 7,000 computing entities across China to accelerate efficient supply-demand matching.

　　Computing power interconnection is making it easier for enterprises to access and utilize resources. “By accessing services provided via the computing power Internet, enterprises can easily utilize massive computing resources. With nationwide access to interconnected computing nodes, tasks such as layout validation, which used to take weeks, can now be completed in just a few days, with efficiency improving by several times,” said a related leader from Daicy Software in Beijing. By leveraging the computing power Internet, enterprises can run multiple simulation tests across various scenarios in parallel. Compared to the previous one-by-one testing approach, efficiency has significantly improved, allowing design flaws to be identified and corrected more quickly. As a result, the success rate of product design has increased from 60% to 80%. Meanwhile, domestically developed computing power continues to make strong strides. “Based on our self-developed heterogeneous many-core architecture accelerators, we have built a self-controlled, high-density intelligent computing base using efficient liquid-cooling technology. Each computing unit integrates 32 high-density 2U 8-card liquid-cooled servers, totaling 256 AI accelerator chips and delivering up to 80 PFLOPS of inference performance per cabinet,” a related leader from Tecorigin shared. Currently, the company has served over 200 universities and enterprises, supporting total computing demands exceeding 1,200 PFLOPS. It has also partnered with the National Supercomputing Center in Wuxi to deploy the upgraded “TaihuLight A+” system, accelerating efforts to make computing resources more accessible.

　　“Computing power interconnection integrates distributed resources to enable intelligent monitoring and on-demand usage, thereby promoting inclusive access to computing power,” said Yu Xiaohui, President of the China Academy of Information and Communications Technology. Computing power can be flexibly purchased and used on a “card-hour” basis, similar to the “kilowatt-hour” billing model used for electricity. This pay-as-you-go model ensures that computing power services are readily accessible to anyone, anytime and anywhere.

　　Optical transmission networks with ultra-high bandwidth and ultra-low latency serve as the “main artery” of the computing power network collaboration. Currently, the 400G optical transmission technology has been put into large-scale commercial use. China Mobile’s 400G all-optical inter-provincial backbone network covers over 135 cities, with some multiplex sections supporting capacities of 800G and 1.2T. China Telecom has completed the world’s first verification of a hybrid 400G and 800G dual-rate network and launched a 400G quantum OTN channel between Hefei and the Inner Mongolia hub. China Unicom has built a full-mesh 400G ROADM backbone network and deployed a 400G/800G all-optical computing network infrastructure in the Greater Bay Area.

　　The computing network brain acts as the “neural center” for computing power scheduling. The three major operators have established a three-tier scheduling architecture at the national, regional, and industry levels to ensure computing resources are visible, dynamically allocated, and efficiently utilized. China Telecom’s “Xirang” platform integrates resources from over 350 cloud pools and uses AI algorithms to optimize scheduling paths, improving cross-domain computing power scheduling efficiency by 60% and raising resource utilization above 75%. China Mobile’s “Computing Network Brain” oversees more than 250,000 network links nationwide and manages 37.6 EFlops of computing power, executing 100 million precise scheduling operations every day. China Unicom’s “Zhishu” platform delivers “efficient in-computing and flexible joint computing” services, bridging isolated computing power resources.

　　From real-time monitoring and fault detection of elevators in old residential communities in Wuxi, Jiangsu Province, to advanced computational experiments in cutting-edge fields like aerospace, earth sciences, marine environments, biomedicine, and industrial manufacturing, the “Sunway TaihuLight” system and its upgraded version, “TaihuLight A+,” continue to fulfill a wide range of computing power needs across China. At the National Supercomputing Center in Wuxi, the super-intelligent integrated computing platform offers enterprises and research institutions full-stack intelligent application computing solutions. These include SWBind, a biological macromolecule 3D structure prediction application for new drug research and development; Zhiqi Assistant, an enterprise-level intelligent assistant deeply integrated with large-scale model technologies; policy-native applications developed on advanced policy models; and large-model-based code development applications that support private deployment. Super-intelligent integration is injecting fresh momentum into the transformation and upgrading of local industries. The coordinated development of supercomputing, intelligent computing, and general-purpose computing is driving the creation of a unified ecosystem that brings together all three forms of computing power across China.

　　Green and Low-Carbon Transformation

　　Accelerating the Integrated Development of Computing Power and Energy

　　Green and low-carbon development is the only path for the computing power industry to overcome energy constraints and achieve large-scale growth. Only by improving energy efficiency and expanding the use of green electricity can the industry break through its high energy consumption bottleneck and ensure improvements in both the quantity and quality of computing power supply.

　　Policy-Driven Integration of Computing Power and Energy: In the eastern region, Beijing’s data centers consumed 534 million kilowatt-hours of green electricity in the first quarter of this year. By implementing the Optimization Plan for Existing Data Centers, Beijing aims to reduce the average annual PUE of these centers to below 1.35 by 2027. In the central region, Hunan is dedicated to developing a green and intelligent computing industry, enhancing the supply of green computing power, and promoting coordinated development between computing power and electricity. In the western region, Qinghai is accelerating the creation of a demonstration zone for the coordinated development of green computing power and electricity, establishing a collaborative model of “clean energy and green computing power” to drive the efficient conversion of green electricity into computing power. Meanwhile, Hohhot in Inner Mongolia invests 128 million yuan annually to build a national hub for green computing power and artificial intelligence.

　　Significant Energy Efficiency Improvement via Advanced Technologies: China Mobile’s Yangtze River Delta (Suzhou) Cloud Computing Center has cut cooling energy consumption by 50% to 60% using liquid cooling technology, reducing the data center’s PUE to below 1.25. Meanwhile, China Unicom’s Sanjiangyuan National Big Data Base in Qinghai has pioneered a collaborative model of “green electricity + green computing power + green finance,” enabling the data center to run entirely on green electricity.

　　China’s computing power industry is at a pivotal stage of growth, advancing in both scale and quality. With the ongoing expansion of computing infrastructure, enhanced computing power supply capacity, and increasingly efficient resource allocation, a greener, smarter ecosystem for advanced computing is rapidly taking shape. This is accelerating the rise of new quality productive forces and injecting stronger digital and intelligent momentum into China’s modernization efforts.